IoannisXiotidis

Bs -> μμ effective lifetime measurement

Introduction

In this section of my twiki the (effective lifetime measurement of the Bs->mumu decay) will be presented. The measurement of the effective lifetime has already been performed for the two other CERN experiments (CMS and LHCb) in their last publications for the Br(Bs->mumu) measurement but never for the ATLAS experiment. Theoretically, the measurement of the effective lifetime assumes that only Bh states decay in the SM making the precise measurement of the effective lifetime prompt to new physics.

Analysis Strategy

After an extensive discussion with my supervisor Prof. Alessandro Cerri the analysis strategy has been defined and presented in this section with the corresponding notes in the following 3 images.

The general overview is summarized in the following bullet list and in addition expanded a bit in the comments bellow:

• B+->JpsiK: mu4_mu6_Lxy0_Jpsimumu (trigger): Used as Control Channel
• Bs->mumu: mu4_mu6_Lxy0_Bmumu (trigger): Analysis channel
• Optimize (in BDT and ct) to minimize cτ uncertainty -> BDT cut, ct range
• Branching fraction fit (from Fabio) -> apply to single BDT bin -> remove any dependence on ε, fs/fd -> convert it into a Nsig fit and not Br
• Apply sPlot on fit result ->
  • ct histograms
  • controls (Background, signal Pt, isolation, etc.)
• Consider fit in ct bins -> mostly as cross check
• Fit MC signal distribution to data signal distribution with 1 parameter (cτ scale + resolution?)

We aim practically to have 3 passes over the above mentioned steps:

1. Pass 1 (Use sPlot)
   • Start with cut including two highest BDT bins
   • ct histogram but also:
     • Signal: Pt, Isolation
     • Background: ct, Pt, isolation
   • Compare with simulations
2. Pass 2 (Explore systematics)
   • Fit systematics (pulls, residuals, etc.)
   • Build ct fit (cτ, cτ spreading parameter)
     • Test on ToyMC -> (cτ, σ) correction
     • Bias ? (Pulls, residuals, low statistics)
3. Pass 3 (Full analysis)
   • Fit (m, cτ) in Data
   • Compare with MC -> validation
   • Signal: m, cτ uncorrelated
   • Background: ? -> probably not we need to check

1) Used Channels

1. B+->JpsiK: used as control channel for the analysis -> the used trigger is the mu4_mu6_Lxy0_Jpsimumu
2. Bs->mumu: used as the analysis channel -> the used trigger is the mu4_mu6_Lxy0_Bmumu

Note: In those two channels the "luck" that we have two triggers that both have an Lxy0 cut is very fortunate. The reason for this is that the Lxy0 cut introduces a bias in the measurement of the lifetime, but since it's present in both triggers we are able to cancel the effect. To be honest this is not a direct cancelling of the effect since the two decays have a very different topology (the Jpsi muons don't point to the PV as the Bmumu) but it's a good thing to have it

2) Optimization

In order to progress with the analysis it is vital to optimize the cuts on BDT (we are going to use one BDT range since part of the sPlot technique is to not include variables that are dependent to the extracted distribution -> for more information refer to the sPlot paper) and ct (in ct what we will obtain might be ct bins). The suggested up to now way of optimizing the BDT and the ct cuts are by attempting to minimize the cτ uncertainty. The optimization can be either performed in Data or in the combination of the signal and bb->mumuX MC, but it needs still to be defined.

3) Invariant mass fit

For this step our strategy is to benefit from the already existing fit created by F. Tresoldi for the Br(Bs->mumu) measurement in ATLAS. For the analysis using the 2015/2016 data, as we are planning to use, a simultaneous fit of Signal + Background in 4 BDT bins has been implemented which is more complicated than what we are planning to use for the effective lifetime measurement. Initially, we are not exactly sure whether we are going to bin in cτ and therefore it might be that a simultaneous fit is not required. In addition dependence on the parameters like signal efficiency, fs/fd etc. are not required. We need a simple fit that will measure the number of signal events (Nsig) and not the Br(Bs->mumu).

4) cτ plot extraction

Reaching this point of the analysis will bring us to make a decision regarding the method we plan to use for the effective lifetime measurement. The baseline plan at this stage which is also going to be mainly expanded to obtain the first result with the current data, is a technique called sPlot. Using this technique we will be able to extract the cτ histogram along with a set of control histograms that we are going to look at for cross check reasons (like Background: cτ, Pt, isolation. Signal: Pt, isolation etc.). The second possible method would be to perform a simultaneous fit in all the cτ bins that we created. The latter method mentioned is more complicated and dependent on the time we will have might be used as a cross check to the first of the two methods.

5) Extract result

A very different method than the two described in the papers of the two other experiments (CMS and LHCb) is planned to be used for the final fit. Instead of creating a fit model and perform a binned maximum likelihood fir on the cτ histogram obtained from the sPlot technique. We are planning to fir the MC signal distribution to the data signal distribution with 1 parameter (cτ scale + resolution). The inclusion of the resolution depends on the dependence of the resolution with the background (mainly). Multiple ways to evaluate this dependence have been discussed (usage of ToyMC).

Dataset and variables

In this section we present the datasets for the Bmumu and BpJpsiKplus channels in addition to the related to our analysis variables:

B+->JpsiK+

Main trigger used for this analysis:

1. Up to run number 284484: HLT_mu4_mu6_bJpsimumu -> Main stream
2. From 284484 up to 302956: HLT_mu4_mu6_bJpsimumu_Lxy0 -> Main stream
3. From 302956 and beyound: HLT_mu4_mu6_bJpsimumu_Lxy0_delayed -> Delayed stream

Main stream periods: 2015 (D,E,F,G,H,J), 2016 (A,B,C,D,E,F,G,I,K,L)

Delayed stream periods: 2016 (D,E,F,G,I,K,L)

path for data: /eos/atlas/atlascerngroupdisk/phys-beauty/BsMuMuRun2/BsMuMuRel21Prod1/ntup/skimmed_[MC/data]/[For MC: Channel]/v3/

The used MC sample according to the Bmumu partial Run2 analysis is: DSID: 300404, events: 886359, candidates: 1238672

Ntuple variables in MC samples related to Jpsi:

Jpsi_iso_7_Chi2_5_LoosePt05	Jpsi_a_2D	Jpsi_eta	Jpsi_MUCALC_massErr	Jpsi_properTime	Jpsi_PlngMin3D
Jpsi_DR	Jpsi_a_3D	Jpsi_phi	Jpsi_mass	Jpsi_properTime_err	Jpsi_PlngMax3D
Jpsi_Lxy	Jpsi_Chi2_PVSV_log1Dz	Jpsi_fitChi2	Jpsi_massErr	Jpsi_properTime_sig
Jpsi_Lxy_err	Jpsi_Chi2_PVSV_log2D	Jpsi_fitNDF	Jpsi_minChi2MuonsIPtoAnyPV	Jpsi_PlngMin2D
Jpsi_Lxy_sig	Jpsi_Chi2_PVSV_log3D	Jpsi_MUCALC_mass	Jpsi_pT	Jpsi_PlngMax2D

Ntuple variables in MC samples related to Jpsi Truth:

TRUTH_Jpsi_pT

TRUTH_Jpsi_eta

TRUTH_Jpsi_phi

TRUTH_Jpsi_mass

TRUTH_Jpsi_PDGID

Ntuple variables in MC samples related to Bplus Truth:

TRUTH_B_pT	TRUTH_B_vtx_x
TRUTH_B_eta	TRUTH_B_vtx_y
TRUTH_B_phi
TRUTH_B_mass
TRUTH_B_PDGID

Jpsi studies for variables

In this section we are going to show the studies performed on the Jpsi both in data and MC for checking the above mentioned variables:

Jpsi masses vs Lxy	Proper decay time distributions calculated with different mass values. The first plot is the proper decay time variable stored in the ntuple. The second is the proper decay time calculated with the MUCALC mass. The third is the proper decay time calculated with the Mass variable. The last one is with the value obtained from the PDG	Bin per bin difference to compare the proper decay time distribution obtained from the variable in the ntuple with the distributions obtained from the different masses in the ntuple. Again Plot 1 is the Ntuple vs MUCALC calculation, Plot 2 is the Ntuple vs Mass, Plot 3 is the Ntuple vs the PDG value	Difference and ratio for all the proper decay time calculations
The MUCALC mass distribution for the 4 highest BDT bins (same BDT bins as Bmumu analysis.	The MUCALC mass distribution for the 4 highest BDT bins taken from the paper pictures. Not very acqurate since it was taken my using an online tool for extracting points which means that clicking the points is not that acquare.	Ratio plot for Ntuple value - Paper value with the uncetainty of clicking the value with low resolution from the picture	Proper decay time at truth level for Bplus, calculated from the PV(x,y) position and the B(x,y) position

Selection of proper decay time calculation

An important question to be asked is which of the available proper decay time calculations to use. In our ntuples 4 possible calculations can be used:

1. Measured Lxy, Measured pT, Measured Mass (with Muon Spectometer = MUCALC)
2. Measured Lxy, Measured pT, Measured Mass (only Inner Detector = VTX)
3. Measured Lxy, Measured pT, Truth Mass (Mass taken from the MC = PDG mass with it's intrinstic distribution + Radiative tails)
4. Measured Lxy, Measured pT, PDG mass (as a constant)

The first check we performed was to see what the properTime variable was using for the calculation. For this it was important to look a per event ratio between the properTime variable from the ntuple and the other available calculations. We performed the excerise on the MC sample for the BpJpsiKp for the Jpsi properTime.

Ratio per event for the proper time calculated with the MUCALC mass in the formula divided by the ntuple properTime variable

Ratio per event for the proper time calculated with the VTX mass in the formula divided by the ntuple properTime variable

Ratio per event for the proper time calculated with the Truth mass in the formula (with a PDG cut to ensure Jpsi masses) divided by the ntuple properTIme variable

Ratio per event for the proper time calculated with the PDG value (without the intrinstic width) in the formula divided by the ntuple properTime variable

from the plots above it can be shown that the ntuple variable properTime is being calculated with the use of the PDG value as a constant. This is a aligned with what was found in the ATHENA code for the derivation however from that side things were not that clear. Additionally, the small differences that are observed were concluded to be due to storage IO of the variable through the progress of the derivation code. Finally we confirmed also by looking in the ntuple making code that the calculation of the proper time in both the Jpsi and the B+ is performed by using the same type of mass.

After the validation now the next step was to compare the four proper decay time calculations with the truth proper decay time from the MC and by taking the difference and look at the resolution to identify the calculation with the best proper time resolution. Since, in the ntuple the truth Jpsi proper time is not there we had to calculated with the rest of the truth variables. For this we assumed that the Jpsi vertex ~ same with the B+ vertex in order to calculate the truth Lxy. The results can be shown in the following three plots:

Difference between the truth Jpsi proper time and the proper time calculated using the measured MUCALC mass. The resolution value is 0.02814

Difference between the truth Jpsi proper time and the proper time calculated using the measured VTX mass. The resolution value is 0.02669

Difference between the truth Jpsi proper time and the proper time calculated using the PDG value for the mass (without the instrinsic width, just the mean value). The resolution value is 0.0282

From the study above it's clear that although the PDG value is the value used to perform the proper time calculation in first place the mass that produces the best resolution is the Jpsi VTX mass. The main reason for this is that the Lxy and pT for the Jpsi are calculated using information from the inner detector (tracks). That means that calculating the proper time with the VTX mass (which uses the same information) we introduce a correlated variable which means that some of the uncertainties are common and therfore cancelled. A confirmation can be seen from the fact that even when adding the MUCALC mass in the calculation the resolution is still better than the constanc PDG mass used in first place since again there is the correlation due to the inner detector information used.

B+ proper time resolution

The above procedure has been also applied for the B+ variables in the ntuple with the following results obtained.

Ratio per even for the proper time calculated with the B MUCALC mass in the formula divide by the ntuple properTime variable

Ratio per even for the proper time calculated with the B VTX mass in the formula divide by the ntuple properTime variable

Ratio per even for the proper time calculated with the B Truth mass in the formula divide by the ntuple properTime variable

Ratio per even for the proper time calculated with the B PDG mass in the formula divide by the ntuple properTime variable

The similar result between the TRUTH and the PDG mass calculation is due to the fact that the B+ mass is generated as a constant value in the MC without any width as the Jpsi. Regarding now the resolution studies the following plots where obtained:

Difference between the truth B proper time and the proper time calculated using the measured MUCALC mass. The resolution value is 0.08531

Difference between the truth B proper time and the proper time calculated using the measured VTX mass. The resolution value is 0.08027

Difference between the truth B proper time and the proper time calculated using the PDG value for the mass. The resolution value is 0.08126

Difference between the truth B proper time and the proper time calculated using the TRUTH value for the mass. The resolution value is 0.08126

Showing again that the VTX mass is the preferable mass to be used for the best proper decay time resolution. The reasons are exactly the same as above for the Jpsi. Finally we see again that PDG and TRUTH values for the resolution match exactly, as expected.

Bs lifetime distribution

In this section the distribution of the lifetime will be shown after applying different set of cuts in the Bs 2015/2016 data ntuple. This study was performed after the Rare decays meeting where the analysis plan was described.

2015/2016 analysis BDT cut (0.416)

Initially we tried to apply the BDT bin 3 BDT cut from the 2015/2016 analysis where in the signal region (SR:5166-5526MeV) we expect only clean Bs signal.

MUCALC mass, VTX mass, Lxy, ProperTime distributions for the Left SB (LSB: 4766-5166 MeV), Signal Region (SR: 5166-5526 MeV) and Right SB (RSB: 5526-5966 MeV), with the BDT Bin 3 cut (0.416)

Proper time distribution in signal region for BDT bin 3

from the plots above it can be seen that the number of events we expect for our proper decay time measurement is very few as expected leading to the result that the main source of uncertainties in the analysis will be the statistical uncertainties.

Rough BDT cut at 0.3

Following the identification of the proper time distribution on a bin where there is clean Bs signal we tried to find a BDT cut which is a good starting point for setting up the analysis procedure. For this reason we added on the top of the previous two plots the same regions with a BDT cut of 0.3 leading to the following results:

MUCALC mass, VTX mass, Lxy and ProperTime distributions for BDT bin 3 of the analysis and a BDT cut of 0.3

Proper time distributions for the BDT bin 3 (RED) and BDT cut 0.3 (Yellow) in the SR of the 2015/2016 dataset

Proper time distribution for the BDT cut 0.3 which is signal + background

from the plots above it can be seen again that the dominant source in the effective lifetime studies is the background which is boosting the number of events in the SR massively. However we know that the signal contribution is going to be small and therefore dominated by statistics.

Bs invariant mass distribution and sPlot

For this study there is a detailed page BsPlot

Other twikis

BsPlot

NumericalLikelihood

Useful links for the analysis

• Git Repository - Gitlab for the analysis

My Links

• ATasteOfTWiki - view a short introductory presentation on TWiki for beginners
• WelcomeGuest - starting points on TWiki
• TWikiUsersGuide - complete TWiki documentation, Quick Start to Reference
• Sandbox - try out TWiki on your own
• IoannisXiotidisSandbox - just for me

My Personal Preferences

• Show tool-tip topic info on mouse-over of WikiWord links, on or off:
  • Set LINKTOOLTIPINFO = off

• Preference for the editor, default is the WYSIWYG editor. The options are raw, wysiwyg:
  • Set EDITMETHOD = wysiwyg

• More preferences TWiki has system wide preferences settings defined in TWikiPreferences. You can customize preferences settings to your needs: To overload a system setting, (1) do a "raw view" on TWikiPreferences, (2) copy a Set VARIABLE = value bullet, (3) do a "raw edit" of your user profile page, (4) add the bullet to the bullet list above, and (5) customize the value as needed. Make sure the settings render as real bullets (in "raw edit", a bullet requires 3 or 6 spaces before the asterisk).

Related Topics

• ChangePassword for changing your password
• ChangeEmailAddress for changing your email address
• UserList has a list of other TWiki users
• UserDocumentationCategory is a list of TWiki user documentation
• UserToolsCategory lists all TWiki user tools